🤗 image matching webui
ONNX-compatible LightGlue: Local Feature Matching at Light Speed. Supports TensorRT, OpenVINO
Multiview matching with deep-learning and hand-crafted local features for COLMAP and other SfM software. Supports high-resolution formats and images with rotations. Both CLI and GUI are supported.
MambaGlue: Fast and Robust Local Feature Matching With Mamba @ ICRA'25
OpenCV图像拼接Pipeline集成 SuperPoint 、LightGlue 特征点检测和匹配深度学习模型
Merge superpoint、lightglue、MixVPR into VINS-FUSION for loop closure with TensorRT
ORB-SLAM3 with SuperPoint, LightGlue, and NetVLAD/CosPlace replacing handcrafted features with learned deep learning components for improved visual SLAM accuracy
SuperPoint and LightGlue with TensorRT. Deploy with C++.
Building a full Visual SLAM pipeline to experiment with different techniques
A cutlass cute implementation of headdim-64 flashattentionv2 TensorRT plugin for LightGlue. Run on Jetson Orin NX 8GB with TensorRT 8.5.2.
Single-image 6DoF Visual Positioning System on top of COLMAP / hloc. CLI + Python API + ROS2 node. Verified on south-building and Cambridge Landmarks ShopFacade.
Make Rendered Database using Rendered Image Obtained by Trained Gaussian Splatting
Research testbed for VLM × photogrammetry. Video/photos → metric-scaled 3D cloud, mesh, and Gaussian splat. ArUco scale, COLMAP SfM+MVS, scene-type-aware pipeline.
Deployable computer vision models optimized for edge inference
Offline explainable computer vision for metal-assembly inspection — C#/.NET 8, Avalonia, OpenCV, DINOv2, FastSAM and LightGlue.
Adaptive re-ranking for VPR using logistic regression to gate expensive geometric matching on retrieval lists.
DepthArt: From Snapshot to Landscape
Finds which source assets appear in a game screenshot, and where: DISK local features, LightGlue matching, RANSAC verification. Qt GUI and headless CLI. Also finds duplicate images.
explores integrating keypoint detection and matching models( with LightGlue)—with OpenCV's DNN module using C++.
Where was this photo taken? Coarse-to-fine geolocation: VLM cold read names the region, Mapillary imagery is fetched on demand, MegaLoc retrieves, LightGlue geometrically verifies -- so it returns 'region only' instead of a wrong pin. Photos + reels. CPU. MIT.
Add a description, image, and links to the lightglue topic page so that developers can more easily learn about it.
To associate your repository with the lightglue topic, visit your repo's landing page and select "manage topics."